COCO Integration

With support from the team behind the COCO dataset, we’ve made it easy to download, visualize, and evaluate on the COCO dataset natively in FiftyOne!

Note

Check out this tutorial to see how you can use FiftyOne to evaluate a model on COCO.

Loading the COCO dataset

The FiftyOne Dataset Zoo provides support for loading both the COCO-2014 and COCO-2017 datasets.

Like all other zoo datasets, you can use load_zoo_dataset() to download and load a COCO split into FiftyOne:

import fiftyone as fo
import fiftyone.zoo as foz

# Download and load the validation split of COCO-2017
dataset = foz.load_zoo_dataset("coco-2017", split="validation")

session = fo.launch_app(dataset)

Note

FiftyOne supports loading annotations for the detection task, including bounding boxes and segmentations.

By default, only the bounding boxes are loaded, but you can customize which label types are loaded via the optional label_types argument (see below for details).

Note

We will soon support loading labels for the keypoints, captions, and panoptic segmentation tasks as well. Stay tuned!

In addition, FiftyOne provides parameters that can be used to efficiently download specific subsets of the COCO dataset, allowing you to quickly explore different slices of the dataset without downloading the entire split.

When performing partial downloads, FiftyOne will use existing downloaded data first if possible before resorting to downloading additional data from the web.

import fiftyone as fo
import fiftyone.zoo as foz

#
# Load 50 random samples from the validation split
#
# Only the required images will be downloaded (if necessary).
# By default, only detections are loaded
#

dataset = foz.load_zoo_dataset(
    "coco-2017",
    split="validation",
    max_samples=50,
    shuffle=True,
)

session = fo.launch_app(dataset)

#
# Load segmentations for 25 samples from the validation split that
# contain cats and dogs
#
# Images that contain all `classes` will be prioritized first, followed
# by images that contain at least one of the required `classes`. If
# there are not enough images matching `classes` in the split to meet
# `max_samples`, only the available images will be loaded.
#
# Images will only be downloaded if necessary
#

dataset = foz.load_zoo_dataset(
    "coco-2017",
    split="validation",
    label_types=["segmentations"],
    classes=["cat", "dog"],
    max_samples=25,
)

session.dataset = dataset

The following parameters are available to configure partial downloads of both COCO-2014 and COCO-2017 by passing them to load_zoo_dataset():

• split (None) and splits (None): a string or list of strings, respectively, specifying the splits to load. Supported values are ("train", "test", "validation"). If neither is provided, all available splits are loaded

• label_types (None): a label type or list of label types to load. Supported values are ("detections", "segmentations"). By default, only detections are loaded

• classes (None): a string or list of strings specifying required classes to load. If provided, only samples containing at least one instance of a specified class will be loaded

• image_ids (None): a list of specific image IDs to load. The IDs can be specified either as <split>/<image-id> strings or <image-id> ints of strings. Alternatively, you can provide the path to a TXT (newline-separated), JSON, or CSV file containing the list of image IDs to load in either of the first two formats

• include_id (False): whether to include the COCO ID of each sample in the loaded labels

• include_license (False): whether to include the COCO license of each sample in the loaded labels, if available. The supported values are:

  • "False" (default): don’t load the license

  • True/"name": store the string license name

  • "id": store the integer license ID

  • "url": store the license URL

• only_matching (False): whether to only load labels that match the classes or attrs requirements that you provide (True), or to load all labels for samples that match the requirements (False)

• num_workers (None): the number of processes to use when downloading individual images. By default, multiprocessing.cpu_count() is used

• shuffle (False): whether to randomly shuffle the order in which samples are chosen for partial downloads

• seed (None): a random seed to use when shuffling

• max_samples (None): a maximum number of samples to load per split. If label_types and/or classes are also specified, first priority will be given to samples that contain all of the specified label types and/or classes, followed by samples that contain at least one of the specified labels types or classes. The actual number of samples loaded may be less than this maximum value if the dataset does not contain sufficient samples matching your requirements

Note

See COCO2017Dataset and COCODetectionDatasetImporter for complete descriptions of the optional keyword arguments that you can pass to load_zoo_dataset().

Loading COCO-formatted data

In addition to loading the COCO datasets themselves, FiftyOne also makes it easy to load your own datasets and model predictions stored in COCO format.

The example code below demonstrates this workflow. First, we generate a JSON file containing COCO-formatted labels to work with:

import os

import fiftyone as fo
import fiftyone.zoo as foz

dataset = foz.load_zoo_dataset("quickstart")

# The directory in which the dataset's images are stored
IMAGES_DIR = os.path.dirname(dataset.first().filepath)

# Export some labels in COCO format
dataset.take(5, seed=51).export(
    dataset_type=fo.types.COCODetectionDataset,
    label_field="ground_truth",
    labels_path="/tmp/coco.json",
)

Now we have a /tmp/coco.json file on disk containing COCO labels corresponding to the images in IMAGES_DIR:

python -m json.tool /tmp/coco.json

{
    "info": {...},
    "licenses": [],
    "categories": [
        {
            "id": 1,
            "name": "airplane",
            "supercategory": null
        },
        ...
    ],
    "images": [
        {
            "id": 1,
            "file_name": "003486.jpg",
            "height": 427,
            "width": 640,
            "license": null,
            "coco_url": null
        },
        ...
    ],
    "annotations": [
        {
            "id": 1,
            "image_id": 1,
            "category_id": 1,
            "bbox": [
                34.34,
                147.46,
                492.69,
                192.36
            ],
            "area": 94773.8484,
            "iscrowd": 0
        },
        ...
    ]
}

We can now use Dataset.from_dir() to load the COCO-formatted labels into a new FiftyOne dataset:

# Load COCO formatted dataset
coco_dataset = fo.Dataset.from_dir(
    dataset_type=fo.types.COCODetectionDataset,
    data_path=IMAGES_DIR,
    labels_path="/tmp/coco.json",
    include_id=True,
)

# COCO categories are also imported
print(coco_dataset.info["categories"])
# [{'id': 1, 'name': 'airplane', 'supercategory': None}, ...]

print(coco_dataset)

Name:        2021.06.28.15.14.38
Media type:  image
Num samples: 5
Persistent:  False
Tags:        []
Sample fields:
    id:               fiftyone.core.fields.ObjectIdField
    filepath:         fiftyone.core.fields.StringField
    tags:             fiftyone.core.fields.ListField(fiftyone.core.fields.StringField)
    metadata:         fiftyone.core.fields.EmbeddedDocumentField(fiftyone.core.metadata.ImageMetadata)
    created_at:       fiftyone.core.fields.DateTimeField
    last_modified_at: fiftyone.core.fields.DateTimeField
    detections:       fiftyone.core.fields.EmbeddedDocumentField(fiftyone.core.labels.Detections)
    coco_id:          fiftyone.core.fields.IntField

In the above call to Dataset.from_dir(), we provide the data_path and labels_path parameters to specify the location of the source images and their COCO labels, respectively, and we set include_id=True so that the COCO ID for each image from our JSON labels will be added to each imported sample.

Note

See COCODetectionDatasetImporter for complete descriptions of the optional keyword arguments that you can pass to Dataset.from_dir().

If your workflow generates model predictions in COCO format, you can use the add_coco_labels() utility method to add them to your dataset as follows:

import fiftyone.utils.coco as fouc

#
# Mock COCO predictions, where:
# - `image_id` corresponds to the `coco_id` field of `coco_dataset`
# - `category_id` corresponds to `coco_dataset.info["categories"]`
#
predictions = [
    {"image_id": 1, "category_id": 2, "bbox": [258, 41, 348, 243], "score": 0.87},
    {"image_id": 2, "category_id": 4, "bbox": [61, 22, 504, 609], "score": 0.95},
]
categories = coco_dataset.info["categories"]

# Add COCO predictions to `predictions` field of dataset
fouc.add_coco_labels(coco_dataset, "predictions", predictions, categories)

# Verify that predictions were added to two images
print(coco_dataset.count("predictions"))  # 2

COCO-style evaluation

By default, evaluate_detections() will use COCO-style evaluation to analyze predictions.

You can also explicitly request that COCO-style evaluation be used by setting the method parameter to "coco".

See this page for more information about using FiftyOne to analyze object detection models.

Note

FiftyOne’s implementation of COCO-style evaluation matches the reference implementation available via pycocotools.

Overview

When running COCO-style evaluation using evaluate_detections():

• Predicted and ground truth objects are matched using a specified IoU threshold (default = 0.50). This threshold can be customized via the iou parameter

• By default, only objects with the same label will be matched. Classwise matching can be disabled via the classwise parameter

• Ground truth objects can have an iscrowd attribute that indicates whether the annotation contains a crowd of objects. Multiple predictions can be matched to crowd ground truth objects. The name of this attribute can be customized by passing the optional iscrowd attribute of COCOEvaluationConfig to evaluate_detections()

When you specify an eval_key parameter, a number of helpful fields will be populated on each sample and its predicted/ground truth objects:

• True positive (TP), false positive (FP), and false negative (FN) counts for the each sample are saved in top-level fields of each sample:

  TP: sample.<eval_key>_tp
  FP: sample.<eval_key>_fp
  FN: sample.<eval_key>_fn
  

• The fields listed below are populated on each individual object instance; these fields tabulate the TP/FP/FN status of the object, the ID of the matching object (if any), and the matching IoU:

  TP/FP/FN: object.<eval_key>
        ID: object.<eval_key>_id
       IoU: object.<eval_key>_iou
  

Note

See COCOEvaluationConfig for complete descriptions of the optional keyword arguments that you can pass to evaluate_detections() when running COCO-style evaluation.

Example evaluation

The example below demonstrates COCO-style detection evaluation on the quickstart dataset from the Dataset Zoo:

import fiftyone as fo
import fiftyone.zoo as foz
from fiftyone import ViewField as F

dataset = foz.load_zoo_dataset("quickstart")
print(dataset)

# Evaluate the objects in the `predictions` field with respect to the
# objects in the `ground_truth` field
results = dataset.evaluate_detections(
    "predictions",
    gt_field="ground_truth",
    method="coco",
    eval_key="eval",
)

# Get the 10 most common classes in the dataset
counts = dataset.count_values("ground_truth.detections.label")
classes = sorted(counts, key=counts.get, reverse=True)[:10]

# Print a classification report for the top-10 classes
results.print_report(classes=classes)

# Print some statistics about the total TP/FP/FN counts
print("TP: %d" % dataset.sum("eval_tp"))
print("FP: %d" % dataset.sum("eval_fp"))
print("FN: %d" % dataset.sum("eval_fn"))

# Create a view that has samples with the most false positives first, and
# only includes false positive boxes in the `predictions` field
view = (
    dataset
    .sort_by("eval_fp", reverse=True)
    .filter_labels("predictions", F("eval") == "fp")
)

# Visualize results in the App
session = fo.launch_app(view=view)

               precision    recall  f1-score   support

       person       0.45      0.74      0.56       783
         kite       0.55      0.72      0.62       156
          car       0.12      0.54      0.20        61
         bird       0.63      0.67      0.65       126
       carrot       0.06      0.49      0.11        47
         boat       0.05      0.24      0.08        37
    surfboard       0.10      0.43      0.17        30
     airplane       0.29      0.67      0.40        24
traffic light       0.22      0.54      0.31        24
        bench       0.10      0.30      0.15        23

    micro avg       0.32      0.68      0.43      1311
    macro avg       0.26      0.54      0.32      1311
 weighted avg       0.42      0.68      0.50      1311

mAP and PR curves

You can compute mean average precision (mAP) and precision-recall (PR) curves for your predictions by passing the compute_mAP=True flag to evaluate_detections():

Note

All mAP calculations are performed according to the COCO evaluation protocol.

import fiftyone as fo
import fiftyone.zoo as foz

dataset = foz.load_zoo_dataset("quickstart")
print(dataset)

# Performs an IoU sweep so that mAP and PR curves can be computed
results = dataset.evaluate_detections(
    "predictions",
    gt_field="ground_truth",
    method="coco",
    compute_mAP=True,
)

print(results.mAP())
# 0.3957

plot = results.plot_pr_curves(classes=["person", "kite", "car"])
plot.show()

Confusion matrices

You can also easily generate confusion matrices for the results of COCO-style evaluations.

In order for the confusion matrix to capture anything other than false positive/negative counts, you will likely want to set the classwise parameter to False during evaluation so that predicted objects can be matched with ground truth objects of different classes.

import fiftyone as fo
import fiftyone.zoo as foz

dataset = foz.load_zoo_dataset("quickstart")

# Perform evaluation, allowing objects to be matched between classes
results = dataset.evaluate_detections(
    "predictions",
    gt_field="ground_truth",
    method="coco",
    classwise=False,
)

# Generate a confusion matrix for the specified classes
plot = results.plot_confusion_matrix(classes=["car", "truck", "motorcycle"])
plot.show()

Note

Did you know? Confusion matrices can be attached to your Session object and dynamically explored using FiftyOne’s interactive plotting features!

mAP protocol

The COCO evaluation protocol is a popular evaluation protocol used by many works in the computer vision community.

COCO-style mAP is derived from VOC-style evaluation with the addition of a crowd attribute and an IoU sweep.

The steps to compute COCO-style mAP are detailed below.

Preprocessing

• Filter ground truth and predicted objects by class (unless classwise=False)

• Sort predicted objects by confidence score so high confidence objects are matched first. Only the top 100 predictions are factored into evaluation (configurable with max_preds)

• Sort ground truth objects so iscrowd objects are matched last

• Compute IoU between every ground truth and predicted object within the same class (and between classes if classwise=False) in each image

• IoU between predictions and crowd objects is calculated as the intersection of both boxes divided by the area of the prediction only. A prediction fully inside the crowd box has an IoU of 1

Matching

Once IoUs have been computed, predictions and ground truth objects are matched to compute true positives, false positives, and false negatives:

• For each class, start with the highest confidence prediction, match it to the ground truth object that it overlaps with the highest IoU. A prediction only matches if the IoU is above the specified iou threshold

• If a prediction matched to a non-crowd object, it will not match to a crowd even if the IoU is higher

• Multiple predictions can match to the same crowd ground truth object, each counting as a true positive

• If a prediction maximally overlaps with a ground truth object that has already been matched (by a higher confidence prediction), the prediction is matched with the next highest IoU ground truth object

• (Only relevant if classwise=False) predictions can only match to crowds if they are of the same class

Computing mAP

• Compute matches for 10 IoU thresholds from 0.5 to 0.95 in increments of 0.05

• The next 6 steps are computed separately for each class and IoU threshold:

• Construct a boolean array of true positives and false positives, sorted (via mergesort) by confidence

• Compute the cumulative sum of the true positive and false positive array

• Compute precision by elementwise dividing the TP-FP-sum array by the total number of predictions up to that point

• Compute recall by elementwise dividing TP-FP-sum array by the number of ground truth objects for the class

• Ensure that precision is a non-increasing array

• Interpolate precision values so that they can be plotted with an array of 101 evenly spaced recall values

• For every class that contains at least one ground truth object, compute the average precision (AP) by averaging the precision values over all 10 IoU thresholds. Then compute mAP by averaging the per-class AP values over all classes